Related Experiment Video
Updated: Feb 19, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.6K
MDPNet: a multi-scale difference perception network for esophageal cancer segmentation in CT images
1Center of Medical Physics, The Second People's Hospital of Changzhou, The Third Affiliated Hospital of Nanjing Medical University, Changzhou, Jiangsu, 213003, China.
BMC Medical Imaging
|February 17, 2026
Summary
A new AI network, MDPNet, improves esophageal cancer segmentation in CT scans. This tool enhances accuracy and generalization for better treatment planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Accurate segmentation of esophageal cancer in CT images is vital for treatment planning.
- Challenges include variable tumor morphology, low contrast, and blurred boundaries.
Purpose of the Study:
- To develop an advanced AI model, MDPNet, for precise esophageal cancer segmentation in CT scans.
- To improve segmentation accuracy and generalization capabilities.
Main Methods:
- Proposed MDPNet (Multi-scale Difference Perception Network) integrating Dynamic Feature Enhancement (DFE), Cross-level Difference Modeling (CDM), and Multi-stage Foreground Enhancement (MFE).
- Evaluated on a self-built ECD 2D dataset and an external test set.
Main Results:
- MDPNet achieved superior performance compared to state-of-the-art methods.
- Reported Dice coefficients of 0.82 on the ECD dataset and 0.78 on the external test set.
- Demonstrated effective improvement in segmentation accuracy and generalization.
Conclusions:
- MDPNet shows significant potential as a decision-support tool for esophageal cancer treatment planning.
- The model exhibits preliminary generalization capability on multi-center test sets.
